Method and system for intelligently detecting construction waste based on unmanned aerial vehicle

Abstract

The invention discloses a method and system for intelligently detecting construction waste based on an unmanned aerial vehicle, and the method comprises the steps: obtaining first inspection image data in a first target task area, and carrying out the first preprocessing of the first inspection image data, and obtaining second inspection image data; presetting a first construction waste detection model, and inputting the second inspection image data into the first construction waste detection model; and carrying out intelligent construction waste detection according to the output of the first construction waste detection model. Intelligent detection of construction waste is performed through the unmanned aerial vehicle, and the detection efficiency and accuracy are greatly improved. Specifically, the unmanned aerial vehicle can comprehensively and quickly acquire the inspection image data in the target task area under the preset path planning, and the quality and availability of the image data are improved through a series of preprocessing steps. And then, intelligent analysis is performed on the processed image data by using a preset detection model, so that efficient recognition of the construction waste is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of drone intelligent detection of construction waste, and in particular to a method and system for drone-based intelligent detection of construction waste. Background Art

[0002] With the continuous advancement of urbanization and the rapid development of the construction industry, the problem of random dumping of construction waste has emerged in an endless stream, seriously affecting the urban environment and the quality of life of residents. The traditional manual inspection method is inefficient, time-consuming and costly, and it is difficult to meet the large-scale and rapidly growing demand for construction waste supervision. In addition, the lack of timely and accurate dynamic monitoring data on the storage of construction waste has undoubtedly increased the difficulty of supervision by urban management departments.

[0003] At present, although drone technology has been widely used in many fields, there are still many challenges and limitations in the detection of construction waste. As the comprehensive utilization rate of construction waste is increased to 60%, market demand has increased sharply. In 2023, the market share of civil drones in my country will reach 72.27%. The application of drone technology in the field of environmental monitoring is not yet mature, especially in the lack of professional products in intelligent garbage inspection. Existing drone detection technology mainly relies on traditional image processing and target detection algorithms. These methods have low recognition accuracy in complex environments and cannot effectively cope with light changes, occlusions and diverse garbage forms. In addition, traditional drone inspections are usually equipped with only a single type of camera, which cannot fully capture and analyze construction waste information in complex terrain and changing environments. These technical limitations have led to unsatisfactory results of existing solutions in practical applications, and a more intelligent and efficient construction waste detection method is urgently needed. Summary of the invention

[0004] The purpose of this section is to summarize some aspects of embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the specification abstract and the invention title of this application to avoid blurring the purpose of this section, the specification abstract and the invention title, and such simplifications or omissions cannot be used to limit the scope of the present invention.

[0005] In view of the above existing problems, the present invention is proposed.

[0006] Therefore, the present invention provides a method and system for intelligently detecting construction waste based on drones, which can solve the problems mentioned in the background technology.

[0007] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0008] In a first aspect, the present invention provides a method for intelligently detecting construction waste based on a drone, comprising:

[0009] Acquire first inspection image data in a first target task area, and perform first preprocessing on the first inspection image data to obtain second inspection image data;

[0010] Presetting a first detection construction waste model, and inputting the second inspection image data into the first detection construction waste model;

[0011] Intelligent detection of construction waste is performed according to the output of the first construction waste detection model.

[0012] As a preferred solution of the method for intelligently detecting construction waste based on drones described in the present invention, wherein: the first inspection image data includes: first-category inspection image data, second-category inspection image data, third-category inspection image data, and fourth-category inspection image data;

[0013] The first inspection image data is all image data acquired when the UAV performs path inspection using a preset first path planning optimization algorithm.

[0014] As a preferred solution of the method for intelligently detecting construction waste based on drones described in the present invention, wherein: the first preprocessing includes at least a first manual preprocessing and a first augmented preprocessing;

[0015] The first manual preprocessing includes a first manual checking operation and a second manual checking operation;

[0016] The first augmentation preprocessing is used to expand the first inspection image data after the first manual preprocessing.

[0017] As a preferred solution of the method for intelligently detecting construction waste based on drones described in the present invention, the first path planning optimization algorithm includes:

[0018] The target garbage points and cleaning sites in the first target task area are regarded as nodes of the graph, the paths between the nodes are regarded as edges, and the weight of the edge represents the distance or time process of the path;

[0019] A node of any graph is set as a starting point, and other nodes are selected according to the planning order of the first path planning optimization algorithm, and all nodes are added to the first unvisited set.

[0020] As a preferred solution of the method for intelligently detecting construction waste based on a drone according to the present invention, all image data acquired by the drone during path inspection by using a preset first path planning optimization algorithm include:

[0021] The drone inspection starts from the starting point and selects other nodes according to the planning sequence of the first path planning optimization algorithm;

[0022] When other nodes in the first unvisited set are visited, the node is deleted from the first unvisited set;

[0023] When the first unvisited set is empty, the drone inspection is completed.

[0024] As a preferred solution of the method for intelligently detecting construction waste based on drones according to the present invention, when other nodes in the first unvisited set are visited, deleting the node from the first unvisited set includes:

[0025] When other nodes in the first unvisited set are visited, determine whether the first type of inspection image data, the second type of inspection image data, the third type of inspection image data, and the fourth type of inspection image data generated by the node for the node meet the preset requirements, and if so, delete the node from the first unvisited set;

[0026] If the conditions are met, the first type of inspection image data, the second type of inspection image data, the third type of inspection image data, and the fourth type of inspection image data at the node are reacquired.

[0027] As a preferred solution of the method for intelligent detection of construction waste based on drones described in the present invention, the first construction waste detection model includes: the input is the second inspection image data, the output is the construction waste recognition probability, the loss function is a composite loss function, and the composite loss function at least includes positioning loss.

[0028] In a second aspect, the present invention provides a system for intelligently detecting construction waste based on a drone, comprising:

[0029] An image data acquisition module, used for acquiring first inspection image data in a first target task area, and performing a first preprocessing on the first inspection image data to obtain second inspection image data;

[0030] A model operation model is used to preset a first detection model of construction waste, and input the second inspection image data into the first detection model of construction waste;

[0031] The result acquisition module is used to perform intelligent detection of construction waste according to the output of the first detection model of construction waste.

[0032] In a third aspect, the present invention provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the above-mentioned method when executing the computer program.

[0033] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program implements the steps of the method described above when executed by a processor.

[0034] Compared with the prior art, the beneficial effects of the present invention are as follows: the present invention proposes a method and system for intelligently detecting construction waste based on unmanned aerial vehicles, obtaining first inspection image data in a first target task area, and performing first preprocessing on the first inspection image data to obtain second inspection image data; presetting a first detection model for construction waste, and inputting the second inspection image data into the first detection model for construction waste; and performing intelligent detection of construction waste according to the output of the first detection model for construction waste. The present invention uses unmanned aerial vehicles to perform intelligent detection of construction waste, which greatly improves the detection efficiency and accuracy. Specifically, under the preset path planning, the unmanned aerial vehicle can comprehensively and quickly obtain the inspection image data in the target task area, and improve the quality and availability of the image data through a series of preprocessing steps. Subsequently, the processed image data is intelligently analyzed using a preset detection model, thereby realizing efficient identification of construction waste. In summary, the present invention provides a new solution for the intelligent detection of construction waste, which has significant technical progress and application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor. Among them:

[0036] Figure 1 A method flow chart of a method and system for intelligently detecting construction waste based on drones provided by one embodiment of the present invention;

[0037] Figure 2 A schematic diagram of an inspection process of a method and system for intelligently detecting construction waste based on drones provided by an embodiment of the present invention;

[0038] Figure 3 A schematic diagram of a PC control terminal for planning a drone flight route based on a method and system for intelligently detecting construction waste using a drone according to an embodiment of the present invention Figure 1 ;

[0039] Figure 4 A schematic diagram of a PC control terminal for planning a drone flight route based on a method and system for intelligently detecting construction waste using a drone according to an embodiment of the present invention Figure 2 ;

[0040] Figure 5 A schematic diagram of an inspection form of a method and system for intelligently detecting construction waste based on drones provided by an embodiment of the present invention;

[0041] Figure 6 A schematic diagram of a mobile terminal inspection process of a method and system for intelligently detecting construction waste based on drones provided by an embodiment of the present invention;

[0042] Figure 7 An internal structural diagram of a computer device for a method and system for intelligently detecting construction waste based on drones provided in one embodiment of the present invention. DETAILED DESCRIPTION

[0043] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0044] Example 1

[0045] Reference Figure 1-Figure 7 , which is the first embodiment of the present invention, provides a method and system for intelligently detecting construction waste based on drones, including:

[0046] Before describing the embodiments of the present application in detail, some related concepts are first explained for the sake of clarity.

[0047] In the related technologies, the detection of construction waste by drones mainly relies on the collection and analysis of image data. However, traditional methods are often limited by the quality and processing efficiency of image data, resulting in inaccurate detection results or a long time consumption.

[0048] This application provides a method that can effectively solve the above-mentioned problems. Next, we will combine multiple embodiments to explain in detail how to implement the method of intelligent detection of construction waste based on drones;

[0049] Figure 1 A method flow chart of a method and system for intelligently detecting construction waste based on drones is shown, including:

[0050] S101, acquiring first inspection image data in a first target task area, and performing first preprocessing on the first inspection image data to obtain second inspection image data;

[0051] In the embodiment of the present application, the first inspection image data includes: first-category inspection image data, second-category inspection image data, third-category inspection image data, and fourth-category inspection image data;

[0052] The first inspection image data is all image data acquired when the UAV performs path inspection using a preset first path planning optimization algorithm.

[0053] Specifically, the first type of inspection image data may be image data of construction waste taken by a visible light camera, the second type of inspection image data may be image data of river sewage taken by a thermal imaging camera and a visible light camera, the third type of inspection image data may be image data of garbage volume obtained by scanning the ground with a laser radar, and the fourth type of inspection image data may be image data of covered garbage taken by a multispectral camera;

[0054] In an optional embodiment, the first type of inspection image data, the second type of inspection image data, the third type of inspection image data, and the fourth type of inspection image data may be any combination of image data acquired by a visible light camera, a thermal imaging camera and a visible light camera, a laser radar, or a multispectral camera. For example, when the first type of inspection image data is image data of covered garbage taken with a multispectral camera, the fourth type of inspection image data may be image data of construction waste taken with a visible light camera.

[0055] In an optional embodiment, the first type of inspection image data, the second type of inspection image data, the third type of inspection image data, and the fourth type of inspection image data can also be other image data that are increased or decreased according to different target effects. For example, when more refined detection is required, night image data taken by an infrared camera (i.e., the fifth type of inspection image data) can be added to capture the thermal radiation characteristics of construction waste at night, thereby further improving the accuracy and comprehensiveness of detection. In addition, considering the impact of changes in ambient light on image quality, an adaptive illumination compensation algorithm can be introduced to dynamically adjust the collected image data to ensure that high-quality inspection image data can be obtained under various lighting conditions.

[0056] In an optional embodiment, if there is no need to consider the covered garbage, the fourth type of inspection image data can be omitted, thereby reducing the workload of data collection and processing. However, in this application, in order to ensure the comprehensiveness of the detection, this type of data is still retained.

[0057] It should be noted that the acquisition of the above-mentioned image data can comprehensively cover different types of construction waste, including but not limited to visible construction waste, river sewage discharge, ground garbage volume and covered garbage, thereby ensuring the accuracy and comprehensiveness of the detection results. At the same time, by introducing a variety of image acquisition equipment and technologies, such as visible light cameras, thermal imaging cameras, lidar and multispectral cameras, the advantages of various image data can be fully utilized, complement each other, and improve the efficiency and accuracy of detection. In addition, according to the needs of different target effects, the type and quantity of image data can be flexibly adjusted to meet the needs of actual detection work. This multi-source data fusion strategy not only improves the flexibility and adaptability of the detection system, but also provides a rich data source for subsequent intelligent analysis.

[0058] In the embodiment of the present application, the first preprocessing includes at least a first manual preprocessing and a first augmentation preprocessing;

[0059] The first manual preprocessing includes a first manual checking operation and a second manual checking operation;

[0060] The first augmentation preprocessing is used to expand the first inspection image data after the first manual preprocessing.

[0061] In the embodiment of the present application, the first manual pre-processing also includes preliminary pre-operations for the first manual checking operation and the second manual checking operation;

[0062] In a preferred embodiment, different preprocessing strategies are adopted according to different types of image data. For example, for the first type of inspection image data taken by a visible light camera, color correction, denoising, and contrast enhancement may need to be performed; for the second type of inspection image data taken by a thermal imaging camera, temperature calibration and temperature field reconstruction may need to be performed; for the third type of inspection image data obtained by laser radar scanning, point cloud filtering and three-dimensional reconstruction may need to be performed; for the fourth type of inspection image data taken by a multispectral camera, spectral unmixing and classification may need to be performed. After these preprocessing steps, the quality and availability of various types of inspection image data are significantly improved, laying a solid foundation for subsequent intelligent detection.

[0063] In the embodiment of the present application, the first manual verification operation is a manual verification operation for construction waste, and the second manual verification operation is a manual verification operation for sewage discharge. For example, the first manual verification operation can specifically be a manual confirmation of the type, quantity and distribution location of construction waste to ensure that the construction waste information in the image data is accurate. The second manual verification operation is to manually check the sewage discharge in the river to determine whether the sewage discharge phenomenon and its severity are accurately reflected in the image data. Although this manual verification operation is time-consuming, it can significantly improve the accuracy of the detection results and ensure the reliability of subsequent analysis.

[0064] In the embodiment of the present application, the first augmentation preprocessing is a data augmentation operation, the purpose of which is to increase the number of training samples. The present application is implemented by performing geometric and color transformation on the first inspection image data after the first manual preprocessing. Specifically, horizontal flipping, vertical flipping, random angle rotation, scaling, translation, noise addition, etc. are applied to generate more training samples to ensure that the model can adapt to different shooting angles, lighting conditions and background changes.

[0065] In an optional embodiment, the first augmentation preprocessing can also be implemented in other ways, such as introducing transfer learning technology, using models that have been trained in other fields as a basis, and fine-tuning the inspection image data to adapt to the special needs of construction waste detection. This method can not only save a lot of training time and computing resources, but also utilize the generalization ability of existing models to improve the overall performance of the detection system. In addition, in order to further improve the effect of data enhancement, it is also possible to combine a variety of data enhancement methods, such as combining geometric transformation with color transformation, or mixing GAN-generated data with real data to achieve better training results. Through these measures, it can be ensured that the trained model can more accurately identify construction waste in various complex scenarios, providing strong support for subsequent governance work.

[0066] Exemplarily, in an embodiment of the present application, images of construction waste from multiple construction sites are collected, totaling about 1,000 images, ensuring that multiple types of waste (such as concrete, wood, metal, etc.) are covered.

[0067] The image annotation tool was used to annotate the location and type of construction waste in each image, generating a training set of 500 annotated samples and a validation set of 100 samples. The first preprocessing was performed to increase the diversity of the dataset. The first augmentation preprocessing process was to horizontally flip 300 images, rotate (random angle) 200 images, and scale 150 images. The training set generated in total reached 1,850 images.

[0068] It should be noted that the first augmentation preprocessing can expand the original 1,000 data samples to 1,850 data samples. This data augmentation technology not only helps the model better learn the various forms and characteristics of construction waste, but also improves the generalization ability of the model in complex environments. During the training process, the model will be exposed to more diverse image data, so that it can better identify construction waste hidden in different scenes.

[0069] In the embodiment of the present application, the first path planning optimization algorithm includes:

[0070] The target garbage points and cleaning sites in the first target task area are regarded as nodes of the graph, the paths between the nodes are regarded as edges, and the weight of the edge represents the distance or time process of the path;

[0071] A node of any graph is set as a starting point, and other nodes are selected according to the planning order of the first path planning optimization algorithm, and all nodes are added to the first unvisited set.

[0072] In the embodiment of the present application, all image data acquired by the drone when performing path inspection using the preset first path planning optimization algorithm includes:

[0073] The drone inspection starts from the starting point and selects other nodes according to the planning sequence of the first path planning optimization algorithm;

[0074] When other nodes in the first unvisited set are visited, the node is deleted from the first unvisited set;

[0075] When the first unvisited set is empty, the drone inspection is completed.

[0076] In an optional embodiment, the planning order of the first path planning optimization algorithm is to select nodes according to distance, and select the node with the smallest distance from the first unvisited set as the next inspection node of the current node, so as to ensure that the drone can inspect along the shortest path and improve the inspection efficiency. For example, if the starting point is a garbage cleaning site located in the city center, the drone will first fly to the garbage point closest to the site and collect the image data of the point. Subsequently, it will continue to select and fly to the next unvisited node closest to the current location, and repeat this process until all garbage points and cleaning sites have been visited. Such a planning order not only ensures that the drone has the shortest path during the inspection process, but also effectively reduces the flight time and energy consumption of the drone, thereby improving the overall inspection efficiency. At the same time, by continuously collecting and analyzing the image data of each node, the drone can also more accurately identify the type and distribution of construction waste, providing strong support for subsequent processing work.

[0077] In an optional embodiment, the planning order of the first path planning optimization algorithm can also be node selection according to weights. The weights can be node weights designed by technicians or users, that is, the weights of target garbage points and cleaning sites. For example, if certain garbage points or cleaning sites have higher priorities or importance, they can be given higher weights. When planning a path, the drone will give priority to these high-weight nodes and fly to these areas for inspections first. This weight design method can flexibly respond to different inspection needs. For example, some areas may require more frequent or priority inspections due to geographical location, environmental influences or human factors. In this way, drones can plan inspection routes more intelligently, ensure that inspection tasks in important areas are given priority, and further improve the pertinence and efficiency of inspections.

[0078] In an embodiment of the present application, when other nodes in the first unvisited set are visited, deleting the node from the first unvisited set includes:

[0079] When other nodes in the first unvisited set are visited, determine whether the first type of inspection image data, the second type of inspection image data, the third type of inspection image data, and the fourth type of inspection image data generated by the node for the node meet the preset requirements, and if so, delete the node from the first unvisited set;

[0080] If the conditions are met, the first type of inspection image data, the second type of inspection image data, the third type of inspection image data, and the fourth type of inspection image data at the node are reacquired.

[0081] Exemplarily, the target garbage points and cleaning sites in the first target task area are regarded as nodes of the graph, the paths between the nodes are regarded as edges, and the weights of the edges represent the distance or time process of the paths;

[0082] Furthermore, a node of any graph is set as a starting point, and other nodes are selected according to the planning order of the first path planning optimization algorithm, and all nodes are added to the first unvisited set.

[0083] Furthermore, the drone inspection starts from the starting point and selects other nodes according to the planning sequence of the first path planning optimization algorithm;

[0084] Furthermore, when other nodes in the first unvisited set are visited, the node is deleted from the first unvisited set;

[0085] Furthermore, when the first unvisited set is an empty set, the drone inspection is completed. Based on the shortest path tree, the optimal cleaning route from the starting point to the end point is generated. It can be expressed as:

[0086] Initialization: d[v] = ∞ (for all nodes v), d[start] = 0

[0087] Update distance: d[v] = min(d[v], d[u] + w(u, v))

[0088] Among them, w(u,v) represents the weight of edge (u,v), d[v] represents the currently known shortest distance from the starting point to node v, and u represents the current node (the node with the smallest distance in the first unvisited set).

[0089] It should be noted that the above steps can ensure that the drone not only has the shortest path during the inspection process, but also can efficiently cover all target garbage points and cleaning sites to avoid missing or repeated inspections. At the same time, by strictly checking the image data of each node, it can ensure that the collected data meets the needs of subsequent analysis and processing, and improve the accuracy and reliability of the overall work. In addition, by generating the optimal cleaning route from the starting point to the end point, it can provide clear guidance for subsequent garbage cleaning work, further improving the efficiency and effectiveness of garbage disposal.

[0090] S102, presetting a first detection construction waste model, and inputting the second inspection image data into the first detection construction waste model;

[0091] In an embodiment of the present application, the first construction waste detection model includes: the input is the second inspection image data, the output is the construction waste recognition probability, the loss function is a composite loss function, and the composite loss function at least includes positioning loss.

[0092] In an optional embodiment, the first construction waste detection model can use YOLOv8 as a basic model, and can also be constructed using other deep learning frameworks or algorithms, such as SSD, Faster R-CNN, etc. However, no matter which framework is selected for model construction, the input of the model must be the second inspection image data, and the output must be the construction waste recognition probability or an output result that can represent a similar construction waste recognition probability effect;

[0093] It should be noted that the purpose of this application is to identify construction waste, so the positioning problem must be considered. Therefore, at least the positioning loss must be included in the composite loss function. Other types of designs can be made according to the specific needs of different staff members. For example, classification loss can be added to more accurately determine the type of construction waste, or regression loss can be added to more accurately predict the volume or area of ​​construction waste. The combination of these loss functions will make the model more comprehensive and accurate in identifying construction waste. In addition, in order to improve the generalization ability and robustness of the model, data enhancement technology can also be used to pre-process the second inspection image data, such as rotation, scaling, cropping, etc., to increase the training sample size of the model and improve its adaptability to different scenes and lighting conditions. Finally, by continuously optimizing and adjusting the structure and parameters of the model, the first detection construction waste model can show higher accuracy and efficiency in practical applications, providing strong support for the intelligent detection and cleaning of construction waste.

[0094] In the embodiment of the present application, YOLOv8 is selected as the basic model, and the weight parameters are initialized. The YOLOv8 model is selected, and the corresponding hyperparameters are set according to the actual application requirements. The setting of the learning rate (Learning Rate, η) determines the pace at which the model updates the weights in each iteration. Usually, the learning rate needs to be debugged based on experience to balance the convergence speed and stability. The batch size (Batch Size, m) determines the number of samples used for each gradient update. A larger batch size will increase memory usage, but it can better estimate the gradient, otherwise it will help capture the local optimum. The number of iterations (Epochs, T) refers to the number of times the entire data set is fully learned by the model, usually set to 1000 to 10000 times, depending on the size and complexity of the data set. In the selection of the optimizer, the Adam optimizer is used, which combines the advantages of momentum and adaptive learning rate, and helps to converge faster.

[0095] Furthermore, the preprocessed image is input into the model for forward propagation and back propagation, and the weight parameters are continuously adjusted to minimize the loss function (such as cross entropy loss).

[0096] Exemplarily, model training is performed through a cycle of forward propagation and back propagation. During forward propagation, the input image passes through each layer of the model convolution operation, and the predicted bounding box position (x, y, w, h) and confidence score p are output. The loss function L is a standard for evaluating the error between the model prediction and the actual annotation. The loss function used in this application includes the positioning loss L loc (using mean square error, MSE Loss), confidence loss L conf (Using binary cross entropy, Binary Cross-EntropyLoss) and classification loss Lcls (Using Binary Cross-Entropy Loss). The overall loss function can be expressed as:

[0097] L=λ loc L loc +λ conf L conf +λ cls L cls

[0098] in:

[0099]

[0100] Among them, x i ,y i represents the predicted bounding box center coordinates, represents the true bounding box center coordinates, w i ,h represents the predicted bounding box width and height, Represents the true bounding box width and height.

[0101]

[0102] Among them, p i Indicates the true confidence (i.e., the binary label of whether the target exists), Indicates the confidence of the prediction, L cls Indicates that L conf Similarly, for the binary cross entropy loss used for multi-class classification, during backpropagation, the weight parameters of the model are adjusted by the gradient descent algorithm to minimize the loss function. loc , conf and cls Represents the weight coefficients of each loss function. These weight coefficients can be adjusted according to actual needs to balance the impact of different loss terms on model training. In practical applications, the selection of these weight coefficients often requires multiple experiments and tuning to achieve the best recognition effect.

[0103] Evaluate model performance on the validation set, and improve model accuracy and generalization ability by adjusting hyperparameters and structural optimization. The validation set is used to evaluate the performance of the model during model training to avoid overfitting problems. When preparing data, this application reserves 20% of the images as a validation set. These images do not overlap with the training set and represent the performance of the model on unknown data. The performance of the model is mainly measured by indicators such as accuracy, mean average precision (mAP, Mean Average Precision), recall, and F1 score. Through these indicators, this application can intuitively evaluate the detection accuracy and stability of the model in different scenarios.

[0104] The trained model is deployed on the cloud server, and the V100 GPU is used to accelerate real-time inference to ensure efficient recognition. The final performance of the model is demonstrated through multiple indicators. For example, in the actual drone inspection process, the model can output the bounding box position, classification results, and confidence scores in real time. By deploying the model on the cloud server and combining the computing power of the V100 GPU, the system can quickly process the images taken by the drone after receiving them and output accurate detection results. Non-maximum suppression (NMS) is used to process multiple overlapping detection boxes, retaining the detection results with the highest confidence, and further improving the accuracy of the system. Through these steps, the effect of the model in construction waste inspection has been significantly improved, ensuring efficient detection and recognition in different environments.

[0105] It should be noted that a deep learning target detection algorithm suitable for construction waste detection is selected, and multiple batches of model training and optimization are carried out. The loss function and accuracy during the training process are monitored to ensure that the model gradually converges on the training set. The accuracy of different models and different batches is compared to ensure that the target accuracy must reach at least 95% to pass.

[0106] It should be noted that the collected data is analyzed for model performance evaluation and optimization, including accuracy and speed. The results of drone intelligent detection are compared one by one with the results of manual detection. In the selected test area, drone intelligent detection and manual detection are carried out simultaneously. At the same time, data is collated during the detection process, and photos with an accuracy rate below 95% are manually reviewed. For special application scenarios, a manual review mechanism will also be introduced. The prediction results of the model are double-confirmed through manual review to ensure the accuracy and reliability of the results.

[0107] It should be noted that after users use the application, this application will continue to collect a large amount of data, divide the data set into multiple subsets, use one subset as the training set, and the remaining subsets as the test set. Through multiple cross-validations, the performance of the algorithm on different data sets will be evaluated. At the same time, manual testing will be carried out from time to time, test samples will be updated, test results will be spot-checked, potential improvement points will be found, and active communication will be carried out with users to collect user feedback and opinions and understand their problems and needs during use.

[0108] S103, performing intelligent detection of construction waste according to the output of the first detection construction waste model.

[0109] In an optional embodiment, when construction waste is detected, the system will trigger an alarm mechanism to notify relevant personnel to deal with it in a timely manner. At the same time, the images and detection results taken by the drone will be transmitted to the cloud database in real time for subsequent analysis and recording. In addition, in order to further improve the accuracy and efficiency of detection, the system will also dynamically adjust the detection parameters and strategies based on historical data and real-time detection results, such as optimizing the prediction algorithm of the bounding box, adjusting the weight coefficient of the loss function, etc. Through such continuous optimization, the system can better adapt to the construction waste detection needs in different environments and scenarios, and improve the intelligence and automation level of detection. In practical applications, this method and system based on drone intelligent detection of construction waste can not only greatly improve detection efficiency, reduce labor costs and labor intensity, but also realize comprehensive and accurate monitoring of construction waste, providing strong support for urban environmental protection and sustainable development.

[0110] In the embodiment of the present application, the UAV flight control is controlled by the stm32f103c8t6 chip solution. It includes GPS, air pressure sensor, and 4g module. It is equipped with a two-axis gimbal using a 5.8G wireless image transmission module, and six-way PWM electric adjustment signal output is used to control the six axes. The PCB uses a four-layer board stack, lithium battery leakage protection, battery charge and discharge management, and power detection circuits, and uses RT9013 dual-channel voltage regulator.

[0111] like Figure 2As shown, it is a schematic diagram of the inspection process of a method and system for intelligent detection of construction waste by drones. Start the task: First, the operator will select the shooting height and area according to the needs 201, and then start the drone to take off automatically 202. Shooting process: After the drone flies to the specified height 203, it uses a visible light camera to shoot 204-1, and uses a thermal imaging camera and a visible light camera to shoot the riverbed sewage discharge situation 204-2. Data acquisition: Then, the drone uses a laser radar to scan the ground to obtain garbage volume data 204-3; and uses a multispectral camera to shoot the covered garbage 204-4. Complete the task: After completing the task, the drone will automatically land 207. Data processing and analysis: The collected data will be processed and analyzed 205, and if the accuracy is lower than the threshold, it will be sent to manual review 206. End: Finally, the workflow ends 208.

[0112] like Figure 3 As shown in the figure, it is the PC control terminal for UAV flight route planning. The green part in the figure is the inspection route of the UAV, and the broken line part is the planning sequence of the first path planning optimization algorithm. Figure 4 For the image data of identified construction waste and river sewage, Figure 5 as well as Figure 6 For the subsequent post-recognition processing flow, images with low model recognition rates are sent to the staff's APP for manual review. The system uses a self-developed path planning algorithm to optimize the garbage cleaning route and sends the planning results to the APP of the cleaner. After the cleaning is completed, the staff confirms it in the APP and then generates a cleaning report.

[0113] This embodiment also provides a system for intelligently detecting construction waste based on drones, including:

[0114] An image data acquisition module is used to acquire first inspection image data in a first target task area, and perform first preprocessing on the first inspection image data to obtain second inspection image data;

[0115] The model operation model is used to preset a first detection construction waste model and input the second inspection image data into the first detection construction waste model;

[0116] The result acquisition module is used to perform intelligent detection of construction waste according to the output of the first detection construction waste model.

[0117] The above-mentioned unit modules may be embedded in or independent of a processor in a computer device in the form of hardware, or may be stored in a memory in a computer device in the form of software, so that the processor can call and execute operations corresponding to the above-mentioned modules.

[0118] This embodiment also provides a computer device, which may be a terminal, and its internal structure diagram may be as follows: Figure 7 As shown. The computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. When the computer program is executed by the processor, a method for intelligent detection of construction waste based on drones is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covered on the display screen, or a key, trackball or touchpad set on the computer device housing, or an external keyboard, touchpad or mouse, etc.

[0119] This embodiment further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented:

[0120] Acquire first inspection image data in a first target task area, and perform first preprocessing on the first inspection image data to obtain second inspection image data;

[0121] Preset a first detection construction waste model, and input the second inspection image data into the first detection construction waste model;

[0122] Intelligent detection of construction waste is performed based on the output of the first detection construction waste model.

[0123] Example 2

[0124] Reference Figure 2-Figure 6 , which is an embodiment of the present invention, provides a method and system for intelligent detection of construction waste based on drones. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through experiments.

[0125] In order to further enhance the realism and credibility of the comparative experiment of the drone construction waste detection project, the following are the experimental results after data expansion and refinement.

[0126] Experimental design: A construction site in a certain city (about 1 square kilometer in area) was selected, covering a variety of construction materials and a variety of garbage types. All experiments were conducted in clear weather to ensure data consistency.

[0127] Experimental groups: UAV inspection group, manual inspection group, and satellite remote sensing group;

[0128] 1. Drone inspection team

[0129] Time: 1 hour

[0130] Data collection: Use drones equipped with the YOLOv8 model to collect and analyze construction waste images in real time.

[0131] Identification results: 472 tons of construction waste were identified, including 260 tons of recyclables, with an accuracy rate of 92%.

[0132] Data processing time: images are uploaded and processed in real time, and data feedback is completed within 8 minutes.

[0133] Image acquisition frequency: About 100 high-resolution images can be acquired per minute.

[0134] 2. Manual inspection team

[0135] Time: 4 days (96 hours)

[0136] Data collection: 6 staff members were divided into 2 groups to conduct on-site inspections and record keeping. Each group inspected for about 8 hours a day.

[0137] Identification results: 385 tons of construction waste were identified, including 180 tons of recyclables, with an accuracy rate of 77%.

[0138] Data processing time: Data recording and input takes an additional 2 hours, extending the overall feedback time to 14 hours.

[0139] Inspection efficiency: Each staff member can only handle about 25 tons of construction waste per day.

[0140] 3. Satellite Remote Sensing Group

[0141] Time: 1 day (24 hours, image analysis pending)

[0142] Data collection: Rely on commercial satellites to capture and download images with a maximum resolution of 30cm.

[0143] Identification results: 310 tons of construction waste were identified, including 140 tons of recyclables, with an accuracy rate of 72% (due to low resolution and post-processing issues).

[0144] Data processing time: Satellite image processing and analysis takes 36 hours, and the overall feedback time is extended to 60 hours.

[0145] Coverage: Satellite data cannot completely cover all areas of the earth’s surface, resulting in missed detections.

[0146] Results comparison summary

[0147] Efficiency: The drone team completed data collection and processing in 1 hour, the manual team took 96 hours, and the satellite team needed 60 hours to acquire and process images.

[0148] Accuracy: The drone’s recognition accuracy was 92%, significantly higher than the 77% for the manual group and 72% for the satellite group.

[0149] Identification volume: The drone group identified 472 tons of construction waste, the manual group identified 385 tons, and the satellite group only identified 310 tons.

[0150] In summary, through the experimental design of the control variable method, it is obvious that the UAV construction waste detection project is superior to traditional manual inspection and satellite remote sensing methods in terms of efficiency, accuracy and identification volume. The UAV system not only improves the inspection efficiency, but also ensures the high quality and real-time feedback of data, showing its great potential and rigid demand in modern construction waste management. These results further verify the application value of UAVs in construction waste detection and have the necessity of promotion and application.

[0151] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

[0152] Those skilled in the art will appreciate that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can adopt the form of complete hardware embodiments, complete software embodiments, or embodiments in combination with software and hardware. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code. The scheme in the embodiments of the present application can be implemented in various computer languages, for example, object-oriented programming language Java and literal scripting language JavaScript, etc.

[0153] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0154] These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0155] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0156] Although the preferred embodiments of the present application have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the present application.

[0157] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to include these modifications and variations.

Claims

1. A method for intelligently detecting construction waste based on drones, characterized in that: include: Acquire first inspection image data in a first target task area, and perform first preprocessing on the first inspection image data to obtain second inspection image data; Preset a first detection construction waste model, and input the second inspection image data into the first detection construction waste model; Intelligent detection of construction waste is performed according to the output of the first construction waste detection model.

2. The method for intelligently detecting construction waste based on drones according to claim 1, characterized in that: The first inspection image data includes: first-category inspection image data, second-category inspection image data, third-category inspection image data, and fourth-category inspection image data; The first inspection image data is all image data acquired when the UAV performs path inspection using a preset first path planning optimization algorithm.

3. The method for intelligently detecting construction waste based on drones as claimed in claim 2, characterized in that: The first preprocessing includes at least a first manual preprocessing and a first augmentation preprocessing; The first manual preprocessing includes a first manual checking operation and a second manual checking operation; The first augmentation preprocessing is used to expand the first inspection image data after the first manual preprocessing.

4. The method for intelligently detecting construction waste based on drones as claimed in claim 3, characterized in that: The first path planning optimization algorithm includes: The target garbage points and cleaning sites in the first target task area are regarded as nodes of the graph, the paths between the nodes are regarded as edges, and the weight of the edge represents the distance or time process of the path; A node of any graph is set as a starting point, and other nodes are selected according to the planning order of the first path planning optimization algorithm, and all nodes are added to the first unvisited set.

5. The method for intelligently detecting construction waste based on drones as claimed in claim 4, characterized in that: All image data acquired by the drone when performing path inspection using the preset first path planning optimization algorithm include: The drone inspection starts from the starting point and selects other nodes according to the planning sequence of the first path planning optimization algorithm; When other nodes in the first unvisited set are visited, the node is deleted from the first unvisited set; When the first unvisited set is empty, the drone inspection is completed.

6. The method for intelligently detecting construction waste based on drones as claimed in claim 5, characterized in that: When other nodes in the first unvisited set are visited, deleting the node from the first unvisited set comprises: When other nodes in the first unvisited set are visited, determine whether the first type of inspection image data, the second type of inspection image data, the third type of inspection image data, and the fourth type of inspection image data generated by the node for the node meet the preset requirements, and if so, delete the node from the first unvisited set; If the conditions are met, the first type of inspection image data, the second type of inspection image data, the third type of inspection image data, and the fourth type of inspection image data at the node are reacquired.

7. The method for intelligently detecting construction waste based on drones as claimed in claim 6, characterized in that: The first construction waste detection model includes: input is the second inspection image data, output is the construction waste recognition probability, the loss function is a composite loss function, and the composite loss function at least includes positioning loss.

8. A system for intelligently detecting construction waste based on drones, characterized in that: include: An image data acquisition module, used for acquiring first inspection image data in a first target task area, and performing a first preprocessing on the first inspection image data to obtain second inspection image data; A model operation model is used to preset a first detection model of construction waste, and input the second inspection image data into the first detection model of construction waste; The result acquisition module is used to perform intelligent detection of construction waste according to the output of the first detection model of construction waste.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images